where is the crashlog?

[f.pap]

I'm experiencing some problems with build 3802: MT5 crashes every couple of days with erros like this in the log:

QQ 3 05:25:01.605 Exception C0000005 at 00007FF7CFD51F88 read to FFFFFFFFFFFFFFFF
CD 3 05:25:01.606 Exception C0000005 at 00007FFDDE40E55C read to FFFFFFFFFFFFFFFF
EQ 2 05:25:01.606 Terminal crashlog generated

I'm not using any external DLL, and MT5 also crashes if I only run strategies I compiled many weeks ago and used for a long time without any problem on previous versions of MT5.

Anyway, for better understanding what happened, I would like to have a look at this crashlog file, but I've not been able to find it. Any idea on where it is?

Thanks in advance!

[Fernando Carreiro]

f.pap: I would like to have a look at this crashlog file, but I've not been able to find it. Any idea on where it is?

%APPDATA%\MetaQuotes\Crashes

[f.pap]

Fernando Carreiro #:

%APPDATA%\MetaQuotes\Crashes

Yes, this is the first place I looked at, but this directory is empty. So, it should be somewhere else.

[Fernando Carreiro]

f.pap #: Yes, this is the first place I looked at, but this directory is empty. So, it should be somewhere else.

Forum on trading, automated trading systems and testing trading strategies

What to do with crashlogs?

Slava, 2006.12.18 10:58

Thank You. At restarting client terminal sends crashlog file to the server automatically.

[f.pap]

Does the terminal send the crashlog to MetaQuotes? Is this post still valid? (it dates back to almost 20 years ago and is relative to MT4). Anyway, I guess the file is stored somewhere on the PC, and just would like to find it...

[Alain Verleyen]

Answer from MetaQuotes :

"Crash report can be lost due to critical exception".

[f.pap]

I found out that the crashlog is actually saved in the Crash directory but, when MetaTrader is launched after a crash, it removes this file after (I suppose) sending it to MetaQuotes.

MT5 continues to crash after I updated to build 3802, every couple of days, with my EAs running live (it never crashed during optimizations). I would like to test my EAs with a previous version but apparently it's not possible to downgrade MT5 to a previous version.

I wonder if these crashes are my fault or MT's: I'm not using external DLLs and from what I understood I think that EAs are sandboxed in MT5: I tried to do nasty things with pointers like calling delete of the same pointer twice or dereferencing uninitialized pointers and other things like that, and MT5 never crashed and either ignored the incorrect actions or printed an error in the terminal.

From your experience: is there anything that, if introduced in an MQL5 program, can cause a crash of MT5? In other words, is there any possibility for the code of an EA to cause a crash of MT5?

I would also like to test my EAs with a version of MT5 less than 3800, if there is a way to downgrade to that version.

[Alain Verleyen]

f.pap #:

I found out that the crashlog is actually saved in the Crash directory but, when MetaTrader is launched after a crash, it removes this file after (I suppose) sending it to MetaQuotes.

MT5 continues to crash after I updated to build 3802, every couple of days, with my EAs running live (it never crashed during optimizations). I would like to test my EAs with a previous version but apparently it's not possible to downgrade MT5 to a previous version.

I wonder if these crashes are my fault or MT's: I'm not using external DLLs and from what I understood I think that EAs are sandboxed in MT5: I tried to do nasty things with pointers like calling delete of the same pointer twice or dereferencing uninitialized pointers and other things like that, and MT5 never crashed and either ignored the incorrect actions or printed an error in the terminal.

From your experience: is there anything that, if introduced in an MQL5 program, can cause a crash of MT5? In other words, is there any possibility for the code of an EA to cause a crash of MT5?

I would also like to test my EAs with a version of MT5 less than 3800, if there is a way to downgrade to that version.

A crash is always a bug in the platform which should not crash. However it's in most cases triggered by some code running inside the platform, your EA or maybe an indicator.

You should either try to isolate the portion of code that triggers it, but as I understand that seems hard to reproduce in your case.

I would rather suggest you to upgrade, and not downgrade, as the developers received your crashlog, so they will check it and fix the problem. You can already try Build 3815 by downloading it from this site.

[f.pap]

Alain Verleyen #:

A crash is always a bug in the platform which should not crash. However it's in most cases triggered by some code running inside the platform, your EA or maybe an indicator.

You should either try to isolate the portion of code that triggers it, but as I understand that seems hard to reproduce in your case.

I would rather suggest you to upgrade, and not downgrade, as the developers received your crashlog, so they will check it and fix the problem. You can already try Build 3815 by downloading it from this site.

Ok, I'll try build 3815. Thanks for your reply

[f.pap]

Build 3815 did not solve my problem, I had to wait until September for build 4000. Now, after more than one month without crashes, I can say that my problem is resolved!

I'm very happy, but disappointed because it's not possible to roll back to a previous version: MT5 autonomously decides to update and this can (of course) introduce a new bug. If you just want to try your previous version (which maybe worked perfectly), it has simply gone forever. I started having these problems in June, the solution came in September, with MT5 crashing every one or two days for 3 months!!!

If I could ask a new feature to MetaQuotes, this is the possibility to roll back to a previous version and a better control of the updates.

Now for example, as explained in my new post https://www.mql5.com/en/forum/456223, it seems that these new versions fixed a problem but introduced another bug. Maybe it's my fault, maybe it's an error that I introduced in my code, but if I cannot try on a platform that never showed a problem like this I can only hope and wait for a new update (and a new bug?) or spend hours in looking for a bug that could be not mine.